On-Line Identification of Modal Parameters Using Artificial Neural Networks

Abstract

Using a neural network concept, an algorithm has been developed that can be employed to conduct on-line identification of modal parameters of a structure such as natural frequency, damping ratio, and mode shape coefficients at measurement locations. Using a band-pass filter, the algorithm extracts from the measurement signal the frequency contents in the vicinity of a desired mode. The filtered signal is then used to train a neural network which consists of a linear neuron with three weights. The structure of the neural network has been designed to allow direct identification of modal parameters from the weights and to enhance efficiency for on-line implementation. The algorithm has been implemented on a DSP (digital signal processor) system for performance evaluation. Modal parameter identification tests have been conducted for a laboratory circular plate structure. The algorithm runs approximately at a sampling frequency of up to 50 kHz for the DSP system used for the study and correctly identifies modal properties of the first two modes of the circular plate. This high performance on-line identification algorithm is expected to be useful in improving the performance of control systems that require knowledge of time-varying plant dynamic characteristics.